🔬 physics

Windowed Finite-Size Scaling of Reference-State Information Backflow in Disordered Quantum Reservoirs

This paper introduces a windowed finite-size scaling diagnostic to analyze reference-state information backflow in disordered quantum reservoirs, revealing that finite-reservoir Markovianization is window-dependent and characterized by a disorder-induced transition near $\omega_c \approx 0.41$ that distinguishes boundary-dominated small reservoirs from the scaling behavior of larger extended systems.

Original authors: Mitchell Smith

Published 2026-06-30

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Mitchell Smith

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have a tiny, delicate quantum machine (a two-qubit system) that is supposed to hold a secret. You want to know how long that secret stays safe before it gets lost to the noisy world around it. In the quantum world, information doesn't just vanish; it can sometimes bounce back, like a ball hitting a wall and returning to your hand. This "bouncing back" is called information backflow. When information bounces back, the system remembers its past. When it stops bouncing back and just disappears, the system behaves in a predictable, "Markovian" way (like a coin flip that doesn't care about the previous flip).

This paper asks a simple but tricky question: Does making the "noisy world" (the reservoir) bigger always make the secret disappear faster and more predictably?

The Setup: A Quantum Game of Tag

The researcher, Mitchell Smith, set up a simulation with three main characters:

The Player: A tiny two-qubit system holding a specific quantum state (a "Bell state").
The Crowd (The Reservoir): A chain of other qubits (like a line of people) that the Player is connected to. This crowd is "disordered," meaning the connections between them are random and messy, like a chaotic party.
The Leak: A way for information to escape the crowd entirely, like a hole in the bottom of a bucket.

The goal was to watch how the Player's state changed over time as it interacted with the Crowd. Specifically, the researcher measured how much the Player's state tried to "return" to its original form (the backflow).

The Big Discovery: Size Isn't Everything

Common sense suggests that if you make the Crowd bigger, the Player's secret should disappear faster and more smoothly. You'd expect a straight line: Bigger Crowd = Faster Loss of Memory.

However, the paper found that reality is messier. The relationship depends entirely on which part of the Crowd you are looking at.

The researcher used a clever trick called "Windowed Finite-Size Scaling." Imagine looking at a long line of people through a camera lens.

If you zoom in on the very first few people (the smallest crowds), they act like a small, tight-knit group. They are so small that they act more like a single, coherent unit than a chaotic crowd. In this "window," making the group slightly bigger actually slows down the loss of memory because the group is still too small to be truly chaotic.
If you zoom out to look at a larger group (the "extended" crowd), the behavior changes. Here, the chaos takes over, and making the group bigger does speed up the loss of memory.

The "Crossover" Point

The study found a specific tipping point (around a disorder strength of 0.41) where the behavior flips.

Below the tipping point: The crowd is too orderly. Making it bigger doesn't help it forget; it actually helps it remember longer.
Above the tipping point: The crowd is chaotic enough. Making it bigger helps it forget faster.

The most important finding is that if you try to measure this using the smallest crowds (sizes 2 and 3), you get a distorted picture. It's like trying to judge the weather by looking at a single puddle; you might think it's raining hard because the puddle is deep, but the rest of the street is dry. The smallest reservoirs are "boundary-dominated"—they are too small to represent the true nature of a large, chaotic environment.

The "Leak" Factor

The study also introduced a "leak" (dissipation) into the system. They found that adding a leak suppresses the memory (the backflow) in a predictable, exponential way. They calculated a "Markovianization rate" (a speedometer for how fast the system forgets). This speedometer works well, but only if you look at the right-sized crowd.

The Takeaway

The paper concludes that you cannot understand how a quantum system forgets its past just by counting how big the environment is.

Small environments act like special, memory-holding helpers (auxiliary systems).
Large environments act like statistical drains that wash information away.

To get a true picture of how quantum systems become "classical" (forgetful and irreversible), you have to look at the environment through the right "window." If you look at the wrong size, you might think the system is behaving one way, when it's actually behaving completely differently.

In short: The size of the room matters, but which part of the room you are standing in matters even more. The smallest rooms behave differently than the big halls, and you can't mix them up when trying to understand how the world forgets.

Technical Summary: Windowed Finite-Size Scaling of Reference-State Information Backflow in Disordered Quantum Reservoirs

Problem Statement

The transition from coherent quantum dynamics to effectively irreversible (Markovian) behavior in open quantum systems is a central issue in quantum information and thermodynamics. While it is often assumed that increasing the size of an environment monotonically promotes decoherence and Markovianity, finite structured reservoirs may not follow this trend. Small reservoirs can act as coherent auxiliary systems, while larger ones may exhibit complex internal transport, recurrence, or disorder-dependent memory effects. Current numerical studies often struggle to distinguish between genuine scaling behavior and artifacts arising from minimal reservoir sizes or specific finite-size windows. This work addresses the question of how the route to effective Markovianization depends not only on reservoir size but also on the specific finite-size window used to characterize the scaling.

Methodology

The study employs a controlled numerical protocol involving a two-qubit subsystem coupled to a finite, disordered XY spin-chain reservoir.

System Configuration: The total Hilbert space consists of a subsystem ( $\mathcal{H}_S$ , dimension 4) and a reservoir ( $\mathcal{H}_R$ , dimension $2^N$ ). The subsystem is initialized in the Bell state $|\Phi^+\rangle$ , and the reservoir in the product state $|0\rangle^{\otimes N}$ .
Hamiltonian: The dynamics are governed by $H = H_S + H_R + H_{SR}$ $H = H_{S} + H_{R} + H_{S R}$ .
- $H_S$ is an XY coupling ( $J_S = 0.6$ ).
- $H_R$ includes random local detunings ( $\Omega_r \sim \text{Uniform}(-\omega, \omega)$ ) and nearest-neighbor XY couplings ( $J_r \sim \text{Uniform}(-J_{dis}, J_{dis})$ with $J_{dis}=0.40$ ).
- $H_{SR}$ couples subsystem qubit 1 to all reservoir qubits with strength $g_{SR} = 0.25$ .
Dissipation: Reservoir-local leakage is introduced via dissipative operators $L_r = \sigma_r^-$ with strength $\gamma_{res}$ , evolving the system under a Lindblad master equation.
Diagnostic: Instead of the fully optimized Breuer–Laine–Piilo (BLP) measure, the authors use a reference-state trace-distance backflow diagnostic ( $\mathcal{N}_{ref}$ ). This measures the positive time derivative of the trace distance between the reduced subsystem state $\rho_S(t)$ and a fixed Bell-state reference $|\Phi^-\rangle$ .
Extraction of Markovianization Rate: For each reservoir size $N$ and disorder strength $\omega$ , the dependence of $\mathcal{N}_{ref}$ on leakage $\gamma_{res}$ is fitted to an exponential form $\mathcal{N}_{ref}(\gamma_{res}) = Ae^{-k\gamma_{res}} + C$ . The parameter $k$ is interpreted as an emergent Markovianization rate.
Scaling Analysis: The extracted rates $k(N, \omega)$ are analyzed using windowed finite-size scaling. The authors fit power laws $k \sim N^{\alpha_w}$ over three distinct windows: $N \geq 2$ , $N \geq 3$ , and $N \geq 4$ , to determine if the scaling exponent $\alpha_w$ depends on the inclusion of minimal reservoir sizes.
Simulation Ensemble: The study utilizes a completed ensemble of 34,560 simulations covering reservoir sizes $N \in \{2, 3, 4, 5, 6\}$ , eighteen disorder values in $\omega \in [0.34, 0.48]$ , six leakage strengths, and sixty-four disorder seeds per parameter point.

Key Contributions

Windowed Scaling Diagnostic: The introduction of a method to explicitly compare finite-size scaling windows to separate boundary-dominated regimes (minimal reservoirs) from extended-reservoir scaling behaviors.
Reference-State Backflow Metric: The application of a fixed-reference trace-distance diagnostic to probe recurrence relative to a specific coherent target, offering a reproducible alternative to optimized non-Markovianity measures for studying specific dynamical transitions.
Disorder-Dependent Transition Structure: The identification of a crossover in the scaling exponent of the Markovianization rate as a function of disorder strength, which varies depending on the chosen finite-size window.

Results

Exponential Suppression: The reference-state backflow decreases approximately exponentially with increasing reservoir leakage, allowing for a stable extraction of the Markovianization rate $k$ .
Window-Dependent Scaling:
- Unconstrained Fits ( $N \geq 2$ ): These fits are significantly affected by a minimal-reservoir boundary regime, where the smallest reservoirs behave as boundary-dominated auxiliary systems rather than representative points in a scaling sequence.
- Primary Transition ( $N \geq 2$ analysis): The extracted exponent crosses from negative to positive near a critical disorder $\omega_c \approx 0.417$ . A negative exponent implies that increasing reservoir size reduces the Markovianization rate (memory preservation), while a positive exponent implies that larger reservoirs accelerate the suppression of backflow.
- Extended-Reservoir Window ( $N \geq 3$ ): This window identifies a closely related crossover near $\omega_c \approx 0.411$ .
- Deep Window ( $N \geq 4$ ): This window remains memory-preserving (negative or near-zero scaling) across the sampled disorder range but trends upward with increasing disorder.
Fit Quality Collapse: The quality of power-law fits decreases sharply near the crossover region, indicating competing finite-size scaling tendencies where memory-preserving and Markovianizing behaviors are in conflict.
Phase Map: The heat map of $k(N, \omega)$ reveals that at low disorder, larger reservoirs can reduce the Markovianization rate, whereas at higher disorder, they act as efficient statistical drains.

Significance and Claims

The paper claims that emergent Markovianization in finite structured reservoirs is window-dependent. The transition from coherent finite-reservoir dynamics to effective irreversibility cannot be captured by reservoir size alone; it is contingent upon disorder strength, leakage, and the specific finite-size regime under study.

Operational Irreversibility: The results support an operational view of irreversibility where classical behavior arises not from a breakdown of quantum mechanics, but from the dynamic inaccessibility of information dispersed into environmental degrees of freedom.
Methodological Implication: The study provides a reproducible numerical protocol for separating boundary artifacts from genuine reservoir-size effects. It suggests that the smallest reservoirs in a dataset should not be assumed to be the beginning of a smooth thermodynamic sequence but may require distinct treatment as auxiliary systems.
Limitations: The authors modestly note that the results are model-specific (finite disordered XY chain) and numerical. The diagnostic is reference-based and distinct from the optimized BLP measure. Furthermore, the $N \geq 4$ window in the baseline data contains only three sizes, and the authors note that an extension to $N=7$ is underway to further validate the trends observed in deeper windows.

In summary, the work demonstrates that the route to effective Markovianization is complex and non-monotonic with respect to size in finite systems, necessitating a windowed approach to accurately characterize the scaling of information backflow.